name: gemini-agent-booster version: "2.8.0" description: "Master optimization protocol for Gemini Agent (Antigravity) to unlock native 2M+ long-context reasoning, Gemini 3.x thinking budget control, native context caching, Multimodal Live API protocols, and high-speed problem solving / Protokol optimasi utama untuk Gemini Agent (Antigravity) untuk mengaktifkan pemikiran long-context 2M+, kontrol thinking budget Gemini 3.x, context caching native, protokol Multimodal Live API, dan pemecahan masalah kecepatan tinggi." author: "Roedy Rustam"

Gemini Agent Booster (2026 Edition — Gemini 3.x Ecosystem)

English | Bahasa Indonesia

<a name="english"></a>

English

Orchestration & Integration

Connects and orchestrates with relevant domain skills like brainstorming, zero-to-prod-orchestrator, ai-llm-integration-expert, and project-context-mapper to ensure cohesive execution.

Description

Master optimization protocol for the Gemini Agent (Antigravity) to leverage native Gemini 3.x (Gemini 3.8 Flash, Gemini 3.5/3.1 Pro/Flash) capabilities — including 1M–2M token context window, dynamic thinking budget control, native context caching (cachedContent), Multimodal Live API integration, visual UI auditing, and parallel tool calling.

Trigger Conditions

• Analyzing very large codebases, full log histories, or monolithic documents requiring 1M–2M token context.
• Managing reasoning budgets with Gemini 3.x Thinking Mode (thinkingConfig).
• Implementing cost-saving strategies with native Gemini Context Caching (cachedContent).
• Performing real-time bidirectional multimodal audio/video or visual UI audits.
• Running deep research tasks requiring web search grounding + reasoning synthesis.
• Delegating complex multi-step tasks to parallel agent swarms or browser subagents.

Gemini 3.x Capability Matrix (2026)

1. Dynamic Thinking Budget & Reasoning Protocol

For complex architectural decisions, security audits, or debugging race conditions, control the reasoning depth via thinkingConfig:

• Flash Thinking for Rapid Tasks: Set lower or default thinking budgets for quick bug fixes, linting, and boilerplate generation.
• Extended Thinking for Critical Paths: Allocate high thinking budgets (e.g. 8k–32k thinking tokens) when designing distributed schemas, refactoring core engines, or evaluating cryptographic trade-offs.
• Reasoning Token Separation: Ensure internal thinking tokens (<thought>) are isolated from client-facing output streams so that final responses remain crisp, clean, and token-efficient.

2. Native Context Caching (cachedContent)

Reduce token costs by up to 75–90% and drastically cut latency on large repositories:

• Threshold: Cache prompts, repository snapshots, or API schemas larger than 32,768 tokens.
• TTL Management: Set appropriate time-to-live (TTL, e.g., 1–2 hours for active dev sessions, 24 hours for stable documentation).
• Structure:

```typescript // Native Gemini Context Caching Example const cache = await ai.createCachedContent({ model: 'gemini-3.1-pro', contents: [{ role: 'user', parts: [{ text: fullCodebaseDump }] }], ttl: '3600s', });

const response = await ai.models.generateContent({ model: 'gemini-3.1-pro', contents: [{ role: 'user', parts: [{ text: 'Locate memory leak in worker thread' }] }], cachedContent: cache.name, }); ```

3. 2M+ Token Long-Context Strategies

When inspecting massive codebases:

1. Pass Full File Trees: Use list_dir to obtain the complete project structure, then supply full source files into context.
2. Whole-File Ingestion: With 1M–2M context, avoid grep fragmentation; inspect complete classes and dependency trees in one shot.
3. Cross-Service Traceability: Analyze upstream microservice contracts, protobufs, and frontend consumers concurrently in the same session.
4. Massive Server Logs: Ingest complete production logs to uncover subtle intermittent race conditions and memory leaks.

4. Multimodal Live API & Screen Grounding

Integrate real-time, low-latency multimodal interaction:

• Bidirectional Streaming: Stream audio input and receive audio/text responses over WebSockets using Gemini Multimodal Live API.
• Screen & UI Grounding: Capture frames from browser subagents or desktop windows; Gemini grounds user queries directly to coordinate points on screen.
• Visual UI Auditing Protocol:

1. Capture current running application via browser_subagent.
2. Audit layout, typography, contrast, and visual hierarchy against HIG and WCAG standards.
3. Compare visually with target design using generate_image or design system guidelines.
4. Perform targeted micro-edits to CSS/Tailwind tokens until alignment reaches pixel perfection.

5. Deep Research & Search Grounding

Gemini's native Search Grounding connects the agent directly to real-time web knowledge:

• Use Grounding for fresh library releases, breaking API deprecations, or zero-day CVE lookups.
• Synthesize findings with source attribution citations.

6. Parallel Tool Execution

Gemini natively supports concurrent function calls:

• Read and edit multiple independent files in a single pass.
• Trigger parallel web searches or subagent workers simultaneously to minimize round-trip latency.

<a name="bahasa-indonesia"></a>

Bahasa Indonesia

Integrasi Orkestrasi

Terhubung dan mengorkestrasi skill domain yang relevan seperti brainstorming, zero-to-prod-orchestrator, ai-llm-integration-expert, dan project-context-mapper untuk memastikan eksekusi yang kohesif.

Deskripsi

Protokol optimasi utama untuk Gemini Agent (Antigravity) memanfaatkan kapabilitas ekosistem Gemini 3.x (Gemini 3.8 Flash, Gemini 3.5/3.1 Pro/Flash) — termasuk context window 1M–2M token, kontrol dynamic thinking budget, native context caching (cachedContent), integrasi Multimodal Live API, audit visual UI, dan pemanggilan tool secara paralel.

Kondisi Pemicu

• Menganalisis codebase skala besar, histori log lengkap, atau dokumen monolitik yang membutuhkan konteks 1M–2M token.
• Mengatur alokasi reasoning budget dengan Gemini 3.x Thinking Mode (thinkingConfig).
• Menerapkan strategi pemangkasan biaya melalui native Gemini Context Caching (cachedContent).
• Menjalankan interaksi audio/video dua arah secara real-time atau audit visual UI.
• Menjalankan tugas riset mendalam dengan Google Search Grounding + sintesis penalaran.
• Mendelegasikan tugas multi-langkah ke swarm agen paralel atau browser subagent.

Matriks Kapabilitas Gemini 3.x (2026)

1. Dynamic Thinking Budget & Protokol Penalaran

Untuk keputusan arsitektur kompleks, audit keamanan, atau perbaikan race condition yang rumit, atur kedalaman penalaran via thinkingConfig:

• Flash Thinking untuk Tugas Cepat: Gunakan alokasi budget penalaran default atau rendah untuk perbaikan bug ringan, formatting, dan boilerplate.
• Extended Thinking untuk Alur Kritis: Alokasikan budget penalaran tinggi (misal: 8k–32k thinking tokens) saat mendesain skema terdistribusi, refaktor engine inti, atau evaluasi kriptografi.
• Pemisahan Token Penalaran: Pastikan token pemikiran internal (<thought>) dipisahkan dari aliran output pengguna agar respons akhir tetap ringkas, bersih, dan efisien token.

2. Native Context Caching (cachedContent)

Pangkas biaya API sebesar 75–90% serta kurangi latensi respons pada repositori besar:

• Ambang Batas: Lakukan cache pada prompt, snapshot kode, atau skema dokumen yang melebihi 32.768 token.
• Manajemen TTL: Tetapkan masa aktif cache (misal: 1–2 jam untuk sesi development aktif, 24 jam untuk dokumentasi statis).
• Contoh Implementasi:

``typescript const cache = await ai.createCachedContent({ model: 'gemini-3.1-pro', contents: [{ role: 'user', parts: [{ text: fullCodebaseDump }] }], ttl: '3600s', }); ``

3. Strategi Long-Context 2M+ Token

Saat menganalisis repositori besar:

1. Pohon File Penuh: Gunakan list_dir untuk memetakan struktur proyek, lalu masukkan seluruh file terkait ke dalam konteks.
2. Ingesti File Penuh: Dengan jendela 1M–2M token, hindari fragmentasi grep; periksa seluruh class dan dependency tree secara menyeluruh.
3. Pelacakan Antar Layanan: Analisis kontrak upstream microservice, skema database, dan frontend consumer secara bersamaan dalam satu sesi.
4. Log Produksi Lengkap: Masukkan ratusan ribu baris log untuk mengungkap anomali memori dan race condition intermiten.

4. Multimodal Live API & Screen Grounding

Integrasikan interaksi multimodal latensi rendah secara langsung:

• Streaming Dua Arah: Streaming input suara dan terima respons audio/teks via WebSockets menggunakan Gemini Multimodal Live API.
• Screen & UI Grounding: Tangkap frame layar dari browser subagent; Gemini memetakan perintah ke koordinat visual yang presisi di layar.
• Protokol Audit UI Visual:

1. Ambil screenshot aplikasi yang sedang berjalan via browser_subagent.
2. Audit tata letak, tipografi, kontras, dan konsistensi terhadap pedoman HIG dan WCAG.
3. Bandingkan dengan referensi desain target menggunakan generate_image.
4. Lakukan penyesuaian presisi pada token CSS/Tailwind hingga tampilan mencapai pixel-perfect.

5. Penelitian Mendalam & Search Grounding

Search Grounding native Gemini menghubungkan agen langsung ke informasi web terkini:

• Gunakan Grounding untuk rilis pustaka terbaru, breaking change dokumentasi, atau audit kerentanan CVE terbaru.
• Sintesiskan temuan lengkap dengan sitasi sumber yang dapat diverifikasi.

6. Eksekusi Tool Paralel

Gemini secara native mendukung pemanggilan banyak function call dalam satu giliran (turn):

• Baca dan modifikasi beberapa file independen sekaligus.
• Jalankan pencarian web atau spawn subagent pekerja secara simultan guna meminimalkan total round-trip latency.